As with many legacy systems, the Boeing Fortran 77 math library will be
around for a long time. We have a number of non-portable codes -- machine
constants, date/time, elapsed time, etc. We must have separate versions
for each machine type. In addition, some our our users are invoking
compiler switches such as -i8, -r8.
Our current approach involves generating special codes for each set of
systems and switches. This leads to a configuration control mess.
We would like to create Fortran 90 codes whose interface mimics the old
non-portable code. Then the innards could invoke the proper Fortran 90
capability and the users could link their Fortran 77 codes with the
library we built. Note that the Fortran 90 library would be built with
the same switches the user has.
This ought to work for some vendors who have dropped their Fortran 77
compilers and require the users to invoke a Fortran 90 compiler.
What kind of problems are we likely to get into with this approach. The
systems we have to support include Sun, SGI, HP, IBM RS6000, Cray, and
PCs with Compaq Visual Fortran.
Thanks.
--Stu Anderson
_______________________________________________________________
[log in to unmask] -- Mathematics and Computing Technology
http://www.rt.cs.boeing.com/MEA/comp_math/sla/
Who speaks for Boeing? Not me!
|